Figure 4

Cohort 1. Bar plots of proteins significant for MS versus controls by confidence intervals (95%) applied within group A and within group B. Each bar is means of one protein for each of the four categories: controls in group A (Group A, Ctrl, blue bars, n = 55), MS patients group A (Group A, MS, blue bars, n = 7), controls in group B (Group B, Ctrl, red bars, n = 9), and MS patients in group B (Group B, MS, red bars, n = 30). The protein identities are given in Supplementary Table S2. The y-axes are the abundance levels expressed as z-scores obtained by subtracting means and dividing by standard deviations. (a) Means of the data, (b) means of ER values of group affiliation obtained by ER modelling where impact of MS status is omitted to isolate the effects of group, and (c) means of ER values of MS status obtained by ER modelling where the impact of group affiliation is omitted to isolate effects of disease. The effects of group (panel b) plus the effects of MS status (panel c) gives predicted values (panel a). Importantly, the comparison of MS versus controls when group affiliation is ignored (as in panel a) is dominated by the most frequent categories, which are controls in group A (the blue bars to the left, n = 55) and MS patients in group B (the red bar to the right, n = 30), with the consequence that the lower abundance of these proteins for MS versus controls within group is not observed. Comparisons of MS versus controls based on ER values (displayed in panel c), which isolate the MS-specific effects, revealed the lower expression of these proteins for MS patients compared with controls without confounding impact of group affiliation.